arm: raw_spinlock annotations
Subsystems (7)
cover arm: raw_spinlock annotations
Preempt-RT converts spinlocks to sleeping spinlocks, but some spinlocks need to be kept as real spinlocks. In 2.6.33 we merged the spinlock namespace cleanup which allows us to annotate those locks. The lock type is change to raw_spinlock which is the same as a common spinlock in !RT kernels.
The following series is a collection of lock annotations from the preempt-rt tree. Please merge into 2.6.34.
Thanks,
tglx
PATCH 1/12 arm: Convert dma_spin_lock to raw_spinlock
An embedded and charset-unspecified text was scrubbed...
Name: arm-convert-dma-lock-to-raw-spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/89de0a20/attachment.el>
PATCH 2/12 arm: Convert stop_lock to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-stop-lock-to-raw-spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/8a95639c/attachment.el>
PATCH 3/12 arm-convert-die_lock-undef_lock-to-raw-spinlocks.patch
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-die_lock-undef_lock-to-raw-spinlocks.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/374915af/attachment.el>
PATCH 4/12 arm: Convert l2x0_lock to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-cache-l2x0-lock-to-raw-spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/d8420fc3/attachment.el>
PATCH 5/12 arm: Convert cpu_asid_lock to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-cpu_asid_lock-to-raw_spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/42126210/attachment.el>
PATCH 6/12 arm: Convert xscale minicache_lock to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-xscale-minicache_lock-to-raw_spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/cc2cc271/attachment.el>
PATCH 7/12 arm: Convert v4 minicache_lock to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-v4-minicache_lock-to-raw_spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/0b7908b4/attachment.el>
PATCH 8/12 arm: Convert v6_lock to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-v6_lock-to-raw_spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/b03b9f25/attachment.el>
PATCH 9/12 arm: Convert footbridge gpio and led locks to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-footbridge-gpio-and-led-locks-to-raw_spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/09dbc34c/attachment.el>
PATCH 10/12 arm: Convert integrator cm_lock and v3_lock to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-integrator-locks-to-raw-spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/5b48582e/attachment.el>
PATCH 11/12 arm: Convert ixp4xx pci_lock to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-convert-ixp4xx-pci-lock-to-raw-spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/33b418ff/attachment.el>
PATCH 12/12 arm: Convert shark leds_lock to raw_spinlock
In reply to Thomas GleixnerAn embedded and charset-unspecified text was scrubbed...
Name: arm-shark-convert-leds-lock-to-raw_spinlock.patch
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100217/3b69d64f/attachment.el>
cover arm: raw_spinlock annotations
In reply to Thomas GleixnerHello,
Quoted 8 lines · Thomas Gleixner
Preempt-RT converts spinlocks to sleeping spinlocks, but some spinlocks need to be kept as real spinlocks. In 2.6.33 we merged the spinlock namespace cleanup which allows us to annotate those locks. The lock type is change to
raw_spinlockwhich is the same as a common spinlock in !RT kernels.The following series is a collection of lock annotations from the preempt-rt tree. Please merge into 2.6.34.
Back then I imported these changes in my repo and found them now when cleaning it up. I once tried to get them in[1], and Russell answered:
Thomas asked me about a pull request for his patches, and I said no because they're spread across many different mach-* directories. So this isn't helpful.
I don't understand that critism, as apart from "ARM: footbridge: Convert gpio and led locks to raw_spinlock" most patches (9) only touch a single file, the others only touch a single lock and fix all callers. The footbridge patch admittedly is not that nice, but that's because the touched lock is used all across the tree.
While cleaning up my repo I refound the patches and rebased them on top of today's Linus' tree and only needed to fix up the l2x0_lock patch as in the meantime a new usage hit mainline.
Best regards
Uwe
[1] http://thread.gmane.org/gmane.linux.ports.arm.kernel/74675
Signature · 2 lines
Pengutronix e.K. | Uwe Kleine-K?nig | Industrial Linux Solutions | http://www.pengutronix.de/ |
PATCH 1/12 ARM: Convert dma_spin_lock to raw_spinlock
From: Thomas Gleixner <[email protected]>
dma_spin_lock needs to be a real spinlock on preempt-rt. Convert it to a raw_spinlock. No change for !RT kernels.
--- arch/arm/include/asm/dma.h | 6 +++--- arch/arm/kernel/dma.c | 2 +- 2 files changed, 4 insertions(+), 4 deletions(-)
arch/arm/include/asm/dma.h +3 -3
@@ -31,18 +31,18 @@#define DMA_MODE_CASCADE 0xc0#define DMA_AUTOINIT 0x10 -extern spinlock_t dma_spin_lock; +extern raw_spinlock_t dma_spin_lock;static inline unsigned long claim_dma_lock (void ) {unsigned long flags; -spin_lock_irqsave (&dma_spin_lock, flags); +raw_spin_lock_irqsave (&dma_spin_lock, flags);return flags; }static inline void release_dma_lock (unsigned long flags) { -spin_unlock_irqrestore (&dma_spin_lock, flags); +raw_spin_unlock_irqrestore (&dma_spin_lock, flags); }/* Clear the 'DMA Pointer Flip Flop'.
arch/arm/kernel/dma.c +1 -1
@@ -23,7 +23,7 @@#include <asm/mach/dma.h> -DEFINE_SPINLOCK (dma_spin_lock); +DEFINE_RAW_SPINLOCK (dma_spin_lock);EXPORT_SYMBOL (dma_spin_lock);static dma_t *dma_chan[MAX_DMA_CHANNELS ];
Signature · 1 line
1.7.2.3
PATCH 2/12 ARM: smp: Convert stop_lock to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
stop_lock needs to be a real spinlock on preempt-rt. Convert it to a raw_spinlock. No change for !RT kernels.
--- arch/arm/kernel/smp.c | 6 +++--- 1 files changed, 3 insertions(+), 3 deletions(-)
arch/arm/kernel/smp.c +3 -3
@@ -464,7 +464,7 @@ void __cpuinit percpu_timer_setup(void)local_timer_setup (evt); } -static DEFINE_SPINLOCK (stop_lock); +static DEFINE_RAW_SPINLOCK (stop_lock);/* * ipi_cpu_stop - handle IPI from smp_send_stop()
@@ -473,10 +473,10 @@ static void ipi_cpu_stop(unsigned int cpu) {if (system_state ==SYSTEM_BOOTING || system_state ==SYSTEM_RUNNING ) { -spin_lock (&stop_lock); +raw_spin_lock (&stop_lock);printk (KERN_CRIT "CPU%u: stopping\n" , cpu);dump_stack (); -spin_unlock (&stop_lock); +raw_spin_unlock (&stop_lock); }set_cpu_online (cpu, false);
Signature · 1 line
1.7.2.3
PATCH 3/12 ARM: Convert die_lock and undef_lock to raw_spinlocks
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
die_lock and undef_lock need to be real spinlocks on preempt-rt. Convert them to raw_spinlock. No change for !RT kernels.
Make die_lock static while at it.
--- arch/arm/kernel/traps.c | 20 ++++++++++---------- 1 files changed, 10 insertions(+), 10 deletions(-)
arch/arm/kernel/traps.c +10 -10
@@ -257,7 +257,7 @@ static int __die(const char *str, int err, struct thread_info *thread, struct ptreturn ret; } -DEFINE_SPINLOCK (die_lock); +static DEFINE_RAW_SPINLOCK (die_lock);/* * This function is protected against re-entrancy.
@@ -269,7 +269,7 @@ void die(const char *str, struct pt_regs *regs, int err)oops_enter (); -spin_lock_irq (&die_lock); +raw_spin_lock_irq (&die_lock);console_verbose ();bust_spinlocks (1 ); ret =__die (str, err, thread, regs);
@@ -279,7 +279,7 @@ void die(const char *str, struct pt_regs *regs, int err)bust_spinlocks (0 );add_taint (TAINT_DIE ); -spin_unlock_irq (&die_lock); +raw_spin_unlock_irq (&die_lock);oops_exit ();if (in_interrupt ())
@@ -304,24 +304,24 @@ void arm_notify_die(const char *str, struct pt_regs *regs, }static LIST_HEAD (undef_hook); -static DEFINE_SPINLOCK (undef_lock); +static DEFINE_RAW_SPINLOCK (undef_lock);void register_undef_hook (struct undef_hook *hook) {unsigned long flags; -spin_lock_irqsave (&undef_lock, flags); +raw_spin_lock_irqsave (&undef_lock, flags);list_add (&hook->node, &undef_hook); -spin_unlock_irqrestore (&undef_lock, flags); +raw_spin_unlock_irqrestore (&undef_lock, flags); }void unregister_undef_hook (struct undef_hook *hook) {unsigned long flags; -spin_lock_irqsave (&undef_lock, flags); +raw_spin_lock_irqsave (&undef_lock, flags);list_del (&hook->node); -spin_unlock_irqrestore (&undef_lock, flags); +raw_spin_unlock_irqrestore (&undef_lock, flags); }static int call_undef_hook (struct pt_regs *regs,unsigned int instr)
@@ -330,12 +330,12 @@ static int call_undef_hook(struct pt_regs *regs, unsigned int instr)unsigned long flags;int (*fn)(struct pt_regs *regs,unsigned int instr) =NULL ; -spin_lock_irqsave (&undef_lock, flags); +raw_spin_lock_irqsave (&undef_lock, flags);list_for_each_entry (hook, &undef_hook, node)if ((instr & hook->instr_mask) == hook->instr_val && (regs->ARM_cpsr & hook->cpsr_mask) == hook->cpsr_val) fn = hook->fn; -spin_unlock_irqrestore (&undef_lock, flags); +raw_spin_unlock_irqrestore (&undef_lock, flags);return fn ?fn (regs, instr) :1 ; }
Signature · 1 line
1.7.2.3
PATCH 4/12 ARM: Convert l2x0_lock to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
l2x20_lock needs to be a real spinlock on preempt-rt. Convert it to a raw_spinlock. No change for !RT kernels.
--- arch/arm/mm/cache-l2x0.c | 34 +++++++++++++++++----------------- 1 files changed, 17 insertions(+), 17 deletions(-)
arch/arm/mm/cache-l2x0.c +17 -17
@@ -26,7 +26,7 @@#define CACHE_LINE_SIZE 32static void __iomem *l2x0_base; -static DEFINE_SPINLOCK (l2x0_lock); +static DEFINE_RAW_SPINLOCK (l2x0_lock);static uint32_t l2x0_way_mask;/* Bitmask of active ways */ static inline void cache_wait (void __iomem *reg,unsigned long mask)
@@ -98,9 +98,9 @@ static void l2x0_cache_sync(void) {unsigned long flags; -spin_lock_irqsave (&l2x0_lock, flags); +raw_spin_lock_irqsave (&l2x0_lock, flags);cache_sync (); -spin_unlock_irqrestore (&l2x0_lock, flags); +raw_spin_unlock_irqrestore (&l2x0_lock, flags); }static inline void l2x0_inv_all (void )
@@ -108,11 +108,11 @@ static inline void l2x0_inv_all(void)unsigned long flags;/* invalidate all ways */ -spin_lock_irqsave (&l2x0_lock, flags); +raw_spin_lock_irqsave (&l2x0_lock, flags);writel_relaxed (l2x0_way_mask, l2x0_base +L2X0_INV_WAY );cache_wait (l2x0_base +L2X0_INV_WAY , l2x0_way_mask);cache_sync (); -spin_unlock_irqrestore (&l2x0_lock, flags); +raw_spin_unlock_irqrestore (&l2x0_lock, flags); }static void l2x0_inv_range (unsigned long start,unsigned long end)
@@ -120,7 +120,7 @@ static void l2x0_inv_range(unsigned long start, unsigned long end)void __iomem *base = l2x0_base;unsigned long flags; -spin_lock_irqsave (&l2x0_lock, flags); +raw_spin_lock_irqsave (&l2x0_lock, flags);if (start & (CACHE_LINE_SIZE -1 )) { start &= ~(CACHE_LINE_SIZE -1 );debug_writel (0x03 );
@@ -145,13 +145,13 @@ static void l2x0_inv_range(unsigned long start, unsigned long end) }if (blk_end < end) { -spin_unlock_irqrestore (&l2x0_lock, flags); -spin_lock_irqsave (&l2x0_lock, flags); +raw_spin_unlock_irqrestore (&l2x0_lock, flags); +raw_spin_lock_irqsave (&l2x0_lock, flags); } }cache_wait (base +L2X0_INV_LINE_PA ,1 );cache_sync (); -spin_unlock_irqrestore (&l2x0_lock, flags); +raw_spin_unlock_irqrestore (&l2x0_lock, flags); }static void l2x0_clean_range (unsigned long start,unsigned long end)
@@ -159,7 +159,7 @@ static void l2x0_clean_range(unsigned long start, unsigned long end)void __iomem *base = l2x0_base;unsigned long flags; -spin_lock_irqsave (&l2x0_lock, flags); +raw_spin_lock_irqsave (&l2x0_lock, flags); start &= ~(CACHE_LINE_SIZE -1 );while (start < end) {unsigned long blk_end = start +min (end - start,4096UL );
@@ -170,13 +170,13 @@ static void l2x0_clean_range(unsigned long start, unsigned long end) }if (blk_end < end) { -spin_unlock_irqrestore (&l2x0_lock, flags); -spin_lock_irqsave (&l2x0_lock, flags); +raw_spin_unlock_irqrestore (&l2x0_lock, flags); +raw_spin_lock_irqsave (&l2x0_lock, flags); } }cache_wait (base +L2X0_CLEAN_LINE_PA ,1 );cache_sync (); -spin_unlock_irqrestore (&l2x0_lock, flags); +raw_spin_unlock_irqrestore (&l2x0_lock, flags); }static void l2x0_flush_range (unsigned long start,unsigned long end)
@@ -184,7 +184,7 @@ static void l2x0_flush_range(unsigned long start, unsigned long end)void __iomem *base = l2x0_base;unsigned long flags; -spin_lock_irqsave (&l2x0_lock, flags); +raw_spin_lock_irqsave (&l2x0_lock, flags); start &= ~(CACHE_LINE_SIZE -1 );while (start < end) {unsigned long blk_end = start +min (end - start,4096UL );
@@ -197,13 +197,13 @@ static void l2x0_flush_range(unsigned long start, unsigned long end)debug_writel (0x00 );if (blk_end < end) { -spin_unlock_irqrestore (&l2x0_lock, flags); -spin_lock_irqsave (&l2x0_lock, flags); +raw_spin_unlock_irqrestore (&l2x0_lock, flags); +raw_spin_lock_irqsave (&l2x0_lock, flags); } }cache_wait (base +L2X0_CLEAN_INV_LINE_PA ,1 );cache_sync (); -spin_unlock_irqrestore (&l2x0_lock, flags); +raw_spin_unlock_irqrestore (&l2x0_lock, flags); }void __init l2x0_init (void __iomem *base,__u32 aux_val,__u32 aux_mask)
Signature · 1 line
1.7.2.3
PATCH 5/12 ARM: Convert cpu_asid_lock to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
cpu_asid_lock needs to be a real spinlock on preempt-rt. Convert it to a raw_spinlock. No change for !RT kernels.
--- arch/arm/mm/context.c | 8 ++++---- 1 files changed, 4 insertions(+), 4 deletions(-)
arch/arm/mm/context.c +4 -4
@@ -16,7 +16,7 @@#include <asm/mmu_context.h> #include <asm/tlbflush.h> -static DEFINE_SPINLOCK (cpu_asid_lock); +static DEFINE_RAW_SPINLOCK (cpu_asid_lock);unsigned int cpu_last_asid =ASID_FIRST_VERSION ;#ifdef CONFIG_SMP DEFINE_PER_CPU(struct mm_struct *, current_mm);
@@ -117,7 +117,7 @@ void __new_context(struct mm_struct *mm) {unsigned int asid; -spin_lock (&cpu_asid_lock); +raw_spin_lock (&cpu_asid_lock);#ifdef CONFIG_SMP /* * Check the ASID again, in case the change was broadcast from
@@ -125,7 +125,7 @@ void __new_context(struct mm_struct *mm)*/ if (unlikely (((mm->context.id ^ cpu_last_asid) >>ASID_BITS ) ==0 )) {cpumask_set_cpu (smp_processor_id (),mm_cpumask (mm)); -spin_unlock (&cpu_asid_lock); +raw_spin_unlock (&cpu_asid_lock);return ; }#endif
@@ -153,5 +153,5 @@ void __new_context(struct mm_struct *mm) }set_mm_context (mm, asid); -spin_unlock (&cpu_asid_lock); +raw_spin_unlock (&cpu_asid_lock); }
Signature · 1 line
1.7.2.3
PATCH 6/12 ARM: Convert xscale minicache_lock to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
minicache_lock needs to be a real spinlock on preempt-rt. Convert it to a raw_spinlock. No change for !RT kernels.
--- arch/arm/mm/copypage-xscale.c | 6 +++--- 1 files changed, 3 insertions(+), 3 deletions(-)
arch/arm/mm/copypage-xscale.c +3 -3
@@ -32,7 +32,7 @@#define minicache_pgprot __pgprot(L_PTE_PRESENT | L_PTE_YOUNG | \ L_PTE_MT_MINICACHE) -static DEFINE_SPINLOCK (minicache_lock); +static DEFINE_RAW_SPINLOCK (minicache_lock);/* * XScale mini-dcache optimised copy_user_highpage
@@ -98,14 +98,14 @@ void xscale_mc_copy_user_highpage(struct page *to, struct page *from,if (test_and_clear_bit (PG_dcache_dirty, &from->flags))__flush_dcache_page (page_mapping (from), from); -spin_lock (&minicache_lock); +raw_spin_lock (&minicache_lock);set_pte_ext (TOP_PTE (COPYPAGE_MINICACHE ),pfn_pte (page_to_pfn (from), minicache_pgprot),0 );flush_tlb_kernel_page (COPYPAGE_MINICACHE );mc_copy_user_page ((void *)COPYPAGE_MINICACHE , kto); -spin_unlock (&minicache_lock); +raw_spin_unlock (&minicache_lock);kunmap_atomic (kto,KM_USER1 ); }
Signature · 1 line
1.7.2.3
PATCH 7/12 ARM: Convert v4 minicache_lock to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
minicache_lock needs to be a real spinlock on preempt-rt. Convert it to a raw_spinlock. No change for !RT kernels.
--- arch/arm/mm/copypage-v4mc.c | 6 +++--- 1 files changed, 3 insertions(+), 3 deletions(-)
arch/arm/mm/copypage-v4mc.c +3 -3
@@ -30,7 +30,7 @@#define minicache_pgprot __pgprot(L_PTE_PRESENT | L_PTE_YOUNG | \ L_PTE_MT_MINICACHE) -static DEFINE_SPINLOCK (minicache_lock); +static DEFINE_RAW_SPINLOCK (minicache_lock);/* * ARMv4 mini-dcache optimised copy_user_highpage
@@ -76,14 +76,14 @@ void v4_mc_copy_user_highpage(struct page *to, struct page *from,if (test_and_clear_bit (PG_dcache_dirty, &from->flags))__flush_dcache_page (page_mapping (from), from); -spin_lock (&minicache_lock); +raw_spin_lock (&minicache_lock);set_pte_ext (TOP_PTE (0xffff8000 ),pfn_pte (page_to_pfn (from), minicache_pgprot),0 );flush_tlb_kernel_page (0xffff8000 );mc_copy_user_page ((void *)0xffff8000 , kto); -spin_unlock (&minicache_lock); +raw_spin_unlock (&minicache_lock);kunmap_atomic (kto,KM_USER1 ); }
Signature · 1 line
1.7.2.3
PATCH 8/12 ARM: Convert v6_lock to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
v6_lock needs to be a real spinlock on preempt-rt. Convert it to a raw_spinlock. No change for !RT kernels.
--- arch/arm/mm/copypage-v6.c | 10 +++++----- 1 files changed, 5 insertions(+), 5 deletions(-)
arch/arm/mm/copypage-v6.c +5 -5
@@ -27,7 +27,7 @@#define from_address (0xffff8000)#define to_address (0xffffc000) -static DEFINE_SPINLOCK (v6_lock); +static DEFINE_RAW_SPINLOCK (v6_lock);/* * Copy the user page. No aliasing to deal with so we can just
@@ -89,7 +89,7 @@ static void v6_copy_user_highpage_aliasing(struct page *to,* Now copy the page using the same cache colour as the * pages ultimate destination. */ -spin_lock (&v6_lock); +raw_spin_lock (&v6_lock);set_pte_ext (TOP_PTE (from_address) + offset,pfn_pte (page_to_pfn (from),PAGE_KERNEL ),0 );set_pte_ext (TOP_PTE (to_address) + offset,pfn_pte (page_to_pfn (to),PAGE_KERNEL ),0 );
@@ -102,7 +102,7 @@ static void v6_copy_user_highpage_aliasing(struct page *to,copy_page ((void *)kto, (void *)kfrom); -spin_unlock (&v6_lock); +raw_spin_unlock (&v6_lock); }/*
@@ -122,13 +122,13 @@ static void v6_clear_user_highpage_aliasing(struct page *page, unsigned long vad* Now clear the page using the same cache colour as * the pages ultimate destination. */ -spin_lock (&v6_lock); +raw_spin_lock (&v6_lock);set_pte_ext (TOP_PTE (to_address) + offset,pfn_pte (page_to_pfn (page),PAGE_KERNEL ),0 );flush_tlb_kernel_page (to);clear_page ((void *)to); -spin_unlock (&v6_lock); +raw_spin_unlock (&v6_lock); }struct cpu_user_fns v6_user_fns__initdata = {
Signature · 1 line
1.7.2.3
PATCH 9/12 ARM: footbridge: Convert gpio and led locks to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
nw_gpio_lock and led_lock need to be real spinlocks on preempt-rt. Convert it them to raw_spinlock and fixup all users. No change for !RT kernels.
--- arch/arm/mach-footbridge/include/mach/hardware.h | 2 +- arch/arm/mach-footbridge/netwinder-hw.c | 14 +++++++------- arch/arm/mach-footbridge/netwinder-leds.c | 10 +++++----- drivers/char/ds1620.c | 8 ++++---- drivers/char/nwflash.c | 4 ++-- drivers/mtd/maps/dc21285.c | 4 ++-- sound/oss/waveartist.c | 4 ++-- 7 files changed, 23 insertions(+), 23 deletions(-)
arch/arm/mach-footbridge/include/mach/hardware.h +1 -1
@@ -86,7 +86,7 @@#define CPLD_FLASH_WR_ENABLE 1#ifndef __ASSEMBLY__ -extern spinlock_t nw_gpio_lock; +extern raw_spinlock_t nw_gpio_lock;extern void nw_gpio_modify_op (unsigned int mask,unsigned int set);extern void nw_gpio_modify_io (unsigned int mask,unsigned int in);extern unsigned int nw_gpio_read (void );
arch/arm/mach-footbridge/netwinder-hw.c +7 -7
@@ -68,7 +68,7 @@ static inline void wb977_ww(int reg, int val)/* * This is a lock for accessing ports GP1_IO_BASE and GP2_IO_BASE */ -DEFINE_SPINLOCK (nw_gpio_lock); +DEFINE_RAW_SPINLOCK (nw_gpio_lock);EXPORT_SYMBOL (nw_gpio_lock);static unsigned int current_gpio_op;
@@ -327,9 +327,9 @@ static inline void wb977_init_gpio(void)/* * Set Group1/Group2 outputs */ -spin_lock_irqsave (&nw_gpio_lock, flags); +raw_spin_lock_irqsave (&nw_gpio_lock, flags);nw_gpio_modify_op (-1 ,GPIO_RED_LED |GPIO_FAN ); -spin_unlock_irqrestore (&nw_gpio_lock, flags); +raw_spin_unlock_irqrestore (&nw_gpio_lock, flags); }/*
@@ -390,9 +390,9 @@ static void __init cpld_init(void) {unsigned long flags; -spin_lock_irqsave (&nw_gpio_lock, flags); +raw_spin_lock_irqsave (&nw_gpio_lock, flags);nw_cpld_modify (-1 ,CPLD_UNMUTE |CPLD_7111_DISABLE ); -spin_unlock_irqrestore (&nw_gpio_lock, flags); +raw_spin_unlock_irqrestore (&nw_gpio_lock, flags); } static unsigned char rwa_unlock[] __initdata =
@@ -616,9 +616,9 @@ static int __init nw_hw_init(void)cpld_init ();rwa010_init (); -spin_lock_irqsave (&nw_gpio_lock, flags); +raw_spin_lock_irqsave (&nw_gpio_lock, flags);nw_gpio_modify_op (GPIO_RED_LED |GPIO_GREEN_LED ,DEFAULT_LEDS ); -spin_unlock_irqrestore (&nw_gpio_lock, flags); +raw_spin_unlock_irqrestore (&nw_gpio_lock, flags); }return 0 ; }
arch/arm/mach-footbridge/netwinder-leds.c +5 -5
@@ -31,13 +31,13 @@static char led_state;static char hw_led_state; -static DEFINE_SPINLOCK (leds_lock); +static DEFINE_RAW_SPINLOCK (leds_lock);static void netwinder_leds_event (led_event_t evt) {unsigned long flags; -spin_lock_irqsave (&leds_lock, flags); +raw_spin_lock_irqsave (&leds_lock, flags);switch (evt) {case led_start:
@@ -117,12 +117,12 @@ static void netwinder_leds_event(led_event_t evt)break ; } -spin_unlock_irqrestore (&leds_lock, flags); +raw_spin_unlock_irqrestore (&leds_lock, flags);if (led_state &LED_STATE_ENABLED ) { -spin_lock_irqsave (&nw_gpio_lock, flags); +raw_spin_lock_irqsave (&nw_gpio_lock, flags);nw_gpio_modify_op (GPIO_RED_LED |GPIO_GREEN_LED , hw_led_state); -spin_unlock_irqrestore (&nw_gpio_lock, flags); +raw_spin_unlock_irqrestore (&nw_gpio_lock, flags); } }
drivers/char/ds1620.c +4 -4
@@ -73,21 +73,21 @@ static inline void netwinder_ds1620_reset(void)static inline void netwinder_lock (unsigned long *flags) { -spin_lock_irqsave (&nw_gpio_lock, *flags); +raw_spin_lock_irqsave (&nw_gpio_lock, *flags); }static inline void netwinder_unlock (unsigned long *flags) { -spin_unlock_irqrestore (&nw_gpio_lock, *flags); +raw_spin_unlock_irqrestore (&nw_gpio_lock, *flags); }static inline void netwinder_set_fan (int i) {unsigned long flags; -spin_lock_irqsave (&nw_gpio_lock, flags); +raw_spin_lock_irqsave (&nw_gpio_lock, flags);nw_gpio_modify_op (GPIO_FAN , i ?GPIO_FAN :0 ); -spin_unlock_irqrestore (&nw_gpio_lock, flags); +raw_spin_unlock_irqrestore (&nw_gpio_lock, flags); }static inline int netwinder_get_fan (void )
drivers/char/nwflash.c +2 -2
@@ -618,9 +618,9 @@ static void kick_open(void)* we want to write a bit pattern XXX1 to Xilinx to enable * the write gate, which will be open for about the next 2ms. */ -spin_lock_irqsave (&nw_gpio_lock, flags); +raw_spin_lock_irqsave (&nw_gpio_lock, flags);nw_cpld_modify (CPLD_FLASH_WR_ENABLE ,CPLD_FLASH_WR_ENABLE ); -spin_unlock_irqrestore (&nw_gpio_lock, flags); +raw_spin_unlock_irqrestore (&nw_gpio_lock, flags);/* * let the ISA bus to catch on...
drivers/mtd/maps/dc21285.c +2 -2
@@ -38,9 +38,9 @@ static void nw_en_write(void)* we want to write a bit pattern XXX1 to Xilinx to enable * the write gate, which will be open for about the next 2ms. */ -spin_lock_irqsave (&nw_gpio_lock, flags); +raw_spin_lock_irqsave (&nw_gpio_lock, flags);nw_cpld_modify (CPLD_FLASH_WR_ENABLE ,CPLD_FLASH_WR_ENABLE ); -spin_unlock_irqrestore (&nw_gpio_lock, flags); +raw_spin_unlock_irqrestore (&nw_gpio_lock, flags);/* * let the ISA bus to catch on...
sound/oss/waveartist.c +2 -2
@@ -1483,9 +1483,9 @@ vnc_mute_spkr(wavnc_info *devc) {unsigned long flags; -spin_lock_irqsave (&nw_gpio_lock, flags); +raw_spin_lock_irqsave (&nw_gpio_lock, flags);nw_cpld_modify (CPLD_UNMUTE , devc->spkr_mute_state ?0 :CPLD_UNMUTE ); -spin_unlock_irqrestore (&nw_gpio_lock, flags); +raw_spin_unlock_irqrestore (&nw_gpio_lock, flags); } static void
Signature · 1 line
1.7.2.3
PATCH 10/12 ARM: integrator: Convert cm_lock and v3_lock to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
cm_lock and v3_lock need to be real spinlocks on preempt-rt. Convert them to raw_spinlock. No change for !RT kernels.
--- arch/arm/mach-integrator/core.c | 6 +++--- arch/arm/mach-integrator/pci_v3.c | 14 +++++++------- 2 files changed, 10 insertions(+), 10 deletions(-)
arch/arm/mach-integrator/core.c +3 -3
@@ -203,7 +203,7 @@ static struct amba_pl010_data integrator_uart_data = {#define CM_CTRL IO_ADDRESS(INTEGRATOR_HDR_CTRL) -static DEFINE_SPINLOCK (cm_lock); +static DEFINE_RAW_SPINLOCK (cm_lock);/** * cm_control - update the CM_CTRL register.
@@ -215,10 +215,10 @@ void cm_control(u32 mask, u32 set)unsigned long flags;u32 val; -spin_lock_irqsave (&cm_lock, flags); +raw_spin_lock_irqsave (&cm_lock, flags); val =readl (CM_CTRL ) & ~mask;writel (val | set,CM_CTRL ); -spin_unlock_irqrestore (&cm_lock, flags); +raw_spin_unlock_irqrestore (&cm_lock, flags); }EXPORT_SYMBOL (cm_control);
arch/arm/mach-integrator/pci_v3.c +7 -7
@@ -163,7 +163,7 @@* 7:2 register number * */ -static DEFINE_SPINLOCK (v3_lock); +static DEFINE_RAW_SPINLOCK (v3_lock);#define PCI_BUS_NONMEM_START 0x00000000#define PCI_BUS_NONMEM_SIZE SZ_256M
@@ -284,7 +284,7 @@ static int v3_read_config(struct pci_bus *bus, unsigned int devfn, int where,unsigned long flags;u32 v; -spin_lock_irqsave (&v3_lock, flags); +raw_spin_lock_irqsave (&v3_lock, flags); addr =v3_open_config_window (bus, devfn, where);switch (size) {
@@ -302,7 +302,7 @@ static int v3_read_config(struct pci_bus *bus, unsigned int devfn, int where, }v3_close_config_window (); -spin_unlock_irqrestore (&v3_lock, flags); +raw_spin_unlock_irqrestore (&v3_lock, flags); *val = v;return PCIBIOS_SUCCESSFUL ;
@@ -314,7 +314,7 @@ static int v3_write_config(struct pci_bus *bus, unsigned int devfn, int where,unsigned long addr;unsigned long flags; -spin_lock_irqsave (&v3_lock, flags); +raw_spin_lock_irqsave (&v3_lock, flags); addr =v3_open_config_window (bus, devfn, where);switch (size) {
@@ -335,7 +335,7 @@ static int v3_write_config(struct pci_bus *bus, unsigned int devfn, int where, }v3_close_config_window (); -spin_unlock_irqrestore (&v3_lock, flags); +raw_spin_unlock_irqrestore (&v3_lock, flags);return PCIBIOS_SUCCESSFUL ; }
@@ -510,7 +510,7 @@ void __init pci_v3_preinit(void)hook_fault_code (8 , v3_pci_fault,SIGBUS ,0 ,"external abort on non-linefetch" );hook_fault_code (10 , v3_pci_fault,SIGBUS ,0 ,"external abort on non-linefetch" ); -spin_lock_irqsave (&v3_lock, flags); +raw_spin_lock_irqsave (&v3_lock, flags);/* * Unlock V3 registers, but only if they were previously locked.
@@ -583,7 +583,7 @@ void __init pci_v3_preinit(void)printk (KERN_ERR "PCI: unable to grab PCI error " "interrupt: %d\n" , ret); -spin_unlock_irqrestore (&v3_lock, flags); +raw_spin_unlock_irqrestore (&v3_lock, flags); }void __init pci_v3_postinit (void )
Signature · 1 line
1.7.2.3
PATCH 11/12 ARM: ixp4xx: Convert pci_lock to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
ixp4xx_pci_lock needs to be a real spinlock on preempt-rt. Convert it to a raw_spinlock. No change for !RT kernels.
--- arch/arm/mach-ixp4xx/common-pci.c | 22 +++++++++++----------- 1 files changed, 11 insertions(+), 11 deletions(-)
arch/arm/mach-ixp4xx/common-pci.c +11 -11
@@ -54,7 +54,7 @@ unsigned long ixp4xx_pci_reg_base = 0;* these transactions are atomic or we will end up * with corrupt data on the bus or in a driver. */ -static DEFINE_SPINLOCK (ixp4xx_pci_lock); +static DEFINE_RAW_SPINLOCK (ixp4xx_pci_lock);/* * Read from PCI config space
@@ -62,10 +62,10 @@ static DEFINE_SPINLOCK(ixp4xx_pci_lock);static void crp_read (u32 ad_cbe,u32 *data) {unsigned long flags; -spin_lock_irqsave (&ixp4xx_pci_lock, flags); +raw_spin_lock_irqsave (&ixp4xx_pci_lock, flags); *PCI_CRP_AD_CBE = ad_cbe; *data = *PCI_CRP_RDATA ; -spin_unlock_irqrestore (&ixp4xx_pci_lock, flags); +raw_spin_unlock_irqrestore (&ixp4xx_pci_lock, flags); }/*
@@ -74,10 +74,10 @@ static void crp_read(u32 ad_cbe, u32 *data)static void crp_write (u32 ad_cbe,u32 data) {unsigned long flags; -spin_lock_irqsave (&ixp4xx_pci_lock, flags); +raw_spin_lock_irqsave (&ixp4xx_pci_lock, flags); *PCI_CRP_AD_CBE =CRP_AD_CBE_WRITE | ad_cbe; *PCI_CRP_WDATA = data; -spin_unlock_irqrestore (&ixp4xx_pci_lock, flags); +raw_spin_unlock_irqrestore (&ixp4xx_pci_lock, flags); }static inline int check_master_abort (void )
@@ -101,7 +101,7 @@ int ixp4xx_pci_read_errata(u32 addr, u32 cmd, u32* data)int retval =0 ;int i; -spin_lock_irqsave (&ixp4xx_pci_lock, flags); +raw_spin_lock_irqsave (&ixp4xx_pci_lock, flags); *PCI_NP_AD = addr;
@@ -118,7 +118,7 @@ int ixp4xx_pci_read_errata(u32 addr, u32 cmd, u32* data)if (check_master_abort ()) retval =1 ; -spin_unlock_irqrestore (&ixp4xx_pci_lock, flags); +raw_spin_unlock_irqrestore (&ixp4xx_pci_lock, flags);return retval; }
@@ -127,7 +127,7 @@ int ixp4xx_pci_read_no_errata(u32 addr, u32 cmd, u32* data)unsigned long flags;int retval =0 ; -spin_lock_irqsave (&ixp4xx_pci_lock, flags); +raw_spin_lock_irqsave (&ixp4xx_pci_lock, flags); *PCI_NP_AD = addr;
@@ -140,7 +140,7 @@ int ixp4xx_pci_read_no_errata(u32 addr, u32 cmd, u32* data)if (check_master_abort ()) retval =1 ; -spin_unlock_irqrestore (&ixp4xx_pci_lock, flags); +raw_spin_unlock_irqrestore (&ixp4xx_pci_lock, flags);return retval; }
@@ -149,7 +149,7 @@ int ixp4xx_pci_write(u32 addr, u32 cmd, u32 data)unsigned long flags;int retval =0 ; -spin_lock_irqsave (&ixp4xx_pci_lock, flags); +raw_spin_lock_irqsave (&ixp4xx_pci_lock, flags); *PCI_NP_AD = addr;
@@ -162,7 +162,7 @@ int ixp4xx_pci_write(u32 addr, u32 cmd, u32 data)if (check_master_abort ()) retval =1 ; -spin_unlock_irqrestore (&ixp4xx_pci_lock, flags); +raw_spin_unlock_irqrestore (&ixp4xx_pci_lock, flags);return retval; }
Signature · 1 line
1.7.2.3
PATCH 12/12 ARM: shark: Convert leds_lock to raw_spinlock
In reply to Uwe Kleine-KönigFrom: Thomas Gleixner <[email protected]>
leds_lock needs to be a real spinlock on preempt-rt. Convert it to a raw_spinlock. No change for !RT kernels.
--- arch/arm/mach-shark/leds.c | 6 +++--- 1 files changed, 3 insertions(+), 3 deletions(-)
arch/arm/mach-shark/leds.c +3 -3
@@ -36,7 +36,7 @@ static char led_state;static short hw_led_state;static short saved_state; -static DEFINE_SPINLOCK (leds_lock); +static DEFINE_RAW_SPINLOCK (leds_lock);short sequoia_read (int addr) {outw (addr,0x24 );
@@ -52,7 +52,7 @@ static void sequoia_leds_event(led_event_t evt) {unsigned long flags; -spin_lock_irqsave (&leds_lock, flags); +raw_spin_lock_irqsave (&leds_lock, flags); hw_led_state =sequoia_read (0x09 );
@@ -144,7 +144,7 @@ static void sequoia_leds_event(led_event_t evt)if (led_state &LED_STATE_ENABLED )sequoia_write (hw_led_state,0x09 ); -spin_unlock_irqrestore (&leds_lock, flags); +raw_spin_unlock_irqrestore (&leds_lock, flags); }static int __init leds_init (void )
Signature · 1 line
1.7.2.3
Arnd Bergmann
maintainercover arm: raw_spinlock annotations
In reply to Uwe Kleine-KönigQuoted 3 lines · Uwe Kleine-K?nig
While cleaning up my repo I refound the patches and rebased them on top of today's Linus' tree and only needed to fix up the
l2x0_lockpatch as in the meantime a new usage hit mainline.
The patches all look harmless, but none of them has any information on why the particular locks need to be raw_spin_lock. Ideally a raw spinlock should be the absolute exception, and IMHO should have a comment in front of it why it is special.
Arnd
cover arm: raw_spinlock annotations
Quoted 9 lines · Arnd Bergmann
Quoted 3 lines · Uwe Kleine-K?nig
While cleaning up my repo I refound the patches and rebased them on top of today's Linus' tree and only needed to fix up the
l2x0_lockpatch as in the meantime a new usage hit mainline.The patches all look harmless, but none of them has any information on why the particular locks need to be
raw_spin_lock. Ideally a raw spinlock should be the absolute exception, and IMHO should have a comment in front of it why it is special.
Or at least explained in the patch description.
For instance, can someone explain why the lock for leds and gpio stuff on Footbridge needs to be converted? What is the original problem? More importantly, what is the criteria for using a raw spinlock instead of a normal spinlock?
cover arm: raw_spinlock annotations
Quoted 11 lines · Russell King - ARM Linux
Quoted 9 lines · Arnd Bergmann
Quoted 3 lines · Uwe Kleine-K?nig
While cleaning up my repo I refound the patches and rebased them on top of today's Linus' tree and only needed to fix up the
l2x0_lockpatch as in the meantime a new usage hit mainline.The patches all look harmless, but none of them has any information on why the particular locks need to be
raw_spin_lock. Ideally a raw spinlock should be the absolute exception, and IMHO should have a comment in front of it why it is special.
Quoted 6 lines
Or at least explained in the patch description.
For instance, can someone explain why the lock for leds and gpio stuff on Footbridge needs to be converted? What is the original problem? More importantly, what is the criteria for using a raw spinlock instead of a normal spinlock?
raw_spinlock is still a spinlock when PREEMPT_RT is enabled, mere spinlocks become magically "sleeping" spinlocks (aka. PI aware rtmutexes)
Vs. the patches: IIRC, it was all fallout from running -rt, but that needs to be looked at case by case. Some of those are obvious as they are called deep down in atomic irq disabled code, but others might be just due to laziness reasons.
There is no urgency to get them merged right now.
Thanks,
tglx
Arnd Bergmann
maintainercover arm: raw_spinlock annotations
Quoted 15 lines · On Tuesday 19 October 2010 22:03:32 Thomas Gleixner wrote:
Quoted 13 lines · Russell King - ARM Linux
Quoted 9 lines · Arnd Bergmann
Quoted 3 lines · Uwe Kleine-K?nig
While cleaning up my repo I refound the patches and rebased them on top of today's Linus' tree and only needed to fix up the
l2x0_lockpatch as in the meantime a new usage hit mainline.The patches all look harmless, but none of them has any information on why the particular locks need to be
raw_spin_lock. Ideally a raw spinlock should be the absolute exception, and IMHO should have a comment in front of it why it is special.Or at least explained in the patch description.
Quoted 8 lines
Quoted 4 lines
For instance, can someone explain why the lock for leds and gpio stuff on Footbridge needs to be converted? What is the original problem? More importantly, what is the criteria for using a raw spinlock instead of a normal spinlock?
raw_spinlockis still a spinlock whenPREEMPT_RTis enabled, mere spinlocks become magically "sleeping" spinlocks (aka. PI aware rtmutexes)
I think we all understood that part of the series description.
Quoted 4 lines
Vs. the patches: IIRC, it was all fallout from running -rt, but that needs to be looked at case by case. Some of those are obvious as they are called deep down in atomic irq disabled code, but others might be just due to laziness reasons.
Not as obvious as you'd think. The explanation "called in atomic irq disabled code" makes sense, but I hadn't thought of that before -- I had expected all that code to use IRQ threads in case of PREEMPT_RT.
Requiring a comment there would probably eliminate the laziness issues, adding a comment that you can't be bothered to do the right fix won't help getting a patch merged, so ideally you'd end up with a better solution getting merged.
Arnd
cover arm: raw_spinlock annotations
Quoted 32 lines · Arnd Bergmann
Quoted 15 lines · On Tuesday 19 October 2010 22:03:32 Thomas Gleixner wrote:
Quoted 13 lines · Russell King - ARM Linux
Quoted 9 lines · Arnd Bergmann
Quoted 3 lines · Uwe Kleine-K?nig
While cleaning up my repo I refound the patches and rebased them on top of today's Linus' tree and only needed to fix up the
l2x0_lockpatch as in the meantime a new usage hit mainline.The patches all look harmless, but none of them has any information on why the particular locks need to be
raw_spin_lock. Ideally a raw spinlock should be the absolute exception, and IMHO should have a comment in front of it why it is special.Or at least explained in the patch description.
Quoted 8 lines
Quoted 4 lines
For instance, can someone explain why the lock for leds and gpio stuff on Footbridge needs to be converted? What is the original problem? More importantly, what is the criteria for using a raw spinlock instead of a normal spinlock?
raw_spinlockis still a spinlock whenPREEMPT_RTis enabled, mere spinlocks become magically "sleeping" spinlocks (aka. PI aware rtmutexes)I think we all understood that part of the series description.
Quoted 4 lines
Vs. the patches: IIRC, it was all fallout from running -rt, but that needs to be looked at case by case. Some of those are obvious as they are called deep down in atomic irq disabled code, but others might be just due to laziness reasons.
Quoted 8 lines
Not as obvious as you'd think. The explanation "called in atomic irq disabled code" makes sense, but I hadn't thought of that before -- I had expected all that code to use IRQ threads in case of
PREEMPT_RT.Requiring a comment there would probably eliminate the laziness issues, adding a comment that you can't be bothered to do the right fix won't help getting a patch merged, so ideally you'd end up with a better solution getting merged.
I know, that's why I'm not pressuring that issue. It needs work and looking at the code and context. That few lines can happily live in -rt for now. We have bigger fish to fry there.
Thanks,
tglx